Predicting Inner-Race Bearing Failure via Vibration Sidebands

A cracked inner-race bearing rarely announces itself through temperature. By the time a thermocouple or RTD on a SAG mill pinion bearing housing registers an excursion, the defect has typically progressed through spalling, cage deformation, and secondary damage to the shaft or housing bore. The condition-monitoring problem is therefore not one of detecting hot bearings — it is one of detecting the defect while the raceway is still intact enough to schedule intervention.

This article deals with one failure mode: inner-race bearing cracking. It sets out the mechanism, why conventional temperature alarms miss it, and how envelope-demodulated vibration analysis resolves it — framed against the diagnostic-coverage and proof-test requirements of IEC 61511.

The mechanism: why inner-race cracks stay thermally quiet

A rolling-element bearing carries load across a contact ellipse between the rolling element and the raceway. On the inner ring, that contact zone rotates with the shaft. A subsurface defect — an inclusion, a forging flaw, or a fatigue initiation site — propagates under cyclic contact stress until a crack reaches the surface and a spall forms.

In the early stage, the crack is a stress riser, not a heat source. Friction losses increase marginally. The bearing housing, which may weigh hundreds of kilograms on a SAG mill trunnion or a large conveyor head pulley, acts as a thermal mass that damps any localised temperature rise. A defect that generates intermittent impact energy at the contact zone dissipates that energy as high-frequency structure-borne vibration, not as bulk temperature. Temperature rise becomes measurable only after significant material loss, loss of clearance control, or lubricant breakdown — all late-stage events.

The practical consequence: a temperature-only trip on a critical rotating asset provides diagnostic coverage for late-stage failure, not for incipient inner-race cracking. In IEC 61511 terms, that is a diagnostic coverage gap on the dangerous failure mode, not a sensor fault.

The signature: sidebands around bearing defect frequencies

Inner-race defects produce a characteristic vibration signature. The fundamental is the Ball Pass Frequency Inner (BPFI), a function of shaft speed, bearing geometry, and contact angle. A pure inner-race defect does not appear as a single BPFI peak. Because the defect rotates in and out of the load zone, its amplitude is modulated by shaft rotation. Amplitude modulation produces sidebands spaced at shaft rotational frequency (1X) around the BPFI carrier and its harmonics.

The engineering value is in the sidebands, not the carrier. An inner-race defect generates a family of peaks at BPFI ± n×1X. As the crack propagates, the sideband spacing and amplitude envelope change in a way that correlates with defect severity. Envelope analysis — band-pass filtering around a high-frequency resonance, then demodulating to recover the low-frequency modulation — separates this signature from the broadband noise of a mill or conveyor.

This is where the S5 condition-monitoring design rationale sits: rather than relying on overall velocity RMS or temperature, the analysis targets the demodulated envelope and tracks sideband structure against a baseline captured at commissioning. The output is a trend on defect energy, not a single alarm threshold.

Mapping to IEC 61511 diagnostic coverage and proof-testing

IEC 61511-1 requires that safety instrumented functions be validated against their specified safety integrity, and that diagnostic coverage of dangerous failures be justified. Two points matter here, and both are methodological.

First, a vibration-monitoring channel is not a safety instrumented function and is not a safety device. It is a diagnostic layer that can inform proof-test scope, interval, and inspection targeting. It does not replace the SIF, and no vibration sensor carries a SIL rating by virtue of being a vibration sensor.

Second, where a rotating asset’s dangerous failure mode is bearing seizure or shaft damage, the proof-test regime should account for whether the diagnostic layer can detect the failure mode before it becomes dangerous. A temperature-only proof-test covers late-stage thermal failure. A demodulated vibration channel with a defined detection threshold and a documented response time extends coverage toward incipient inner-race cracking — provided the detection threshold, the sampling interval, and the response time are all specified and verified.

The proof-test implication is direct. If the diagnostic layer can detect an inner-race defect at a defined defect-energy level, and the maintenance response can act within the remaining propagation window, then the interval between intrusive inspections can be justified on evidence rather than on calendar time. That is the closure of the loop: the diagnostic does not guarantee safety, but it provides the evidence that the proof-test interval remains valid.

Practical implementation on mining assets

For a SAG mill pinion or trunnion bearing, mount accelerometers on the bearing housing in the load direction, with a high-frequency resonance above the gear-mesh and mill-rotation harmonics. For a conveyor head pulley or gearbox, the same principle applies, but the modulation source may be the pulley rotation rather than a shaft. For a control valve actuator or a pump, confirm that the BPFI and its sidebands are resolvable above process noise before committing to a detection threshold.

Set the baseline during a known-good run at steady load. Track the demodulated envelope trend, not instantaneous peaks. Define the detection threshold, the confirmation window, and the maintenance action in the same document as the proof-test procedure, so the diagnostic coverage claim is auditable.

Getting started

Identify the rotating assets where inner-race bearing failure is a credible dangerous failure mode and where temperature is currently the only online indicator. For each, confirm bearing geometry to calculate BPFI, capture a baseline envelope spectrum at steady load, and document the sideband structure. Then review the existing proof-test procedure: state the failure mode, the detection method, the threshold, and the response time. Where the diagnostic layer cannot detect incipient inner-race cracking, record that as a coverage gap rather than assuming temperature covers it.

Source: internal-backlog — grounded in S5 engineering design notes, not a current market event.

Talk to S5 System about your site

S5 System designs and manufactures ultra-low-power industrial IoT sensors for mining, agriculture, infrastructure and utilities.

Discuss your IoT monitoring project — book a call.